Challenges and Opportunities of Democracy in the Digital Society

Digital technology is widely perceived to cause important problems, such as fake news, hate speech, and political polarization, which call for policy responses. However, there is no consensus on the specific nature or intensity of those problems nor, therefore, what kinds of actions would be appropriate. On the one hand, there is nothing special about this lack of agreement. The contestation of the nature of problems and solutions is a key feature of politics, in any area. On the other hand, the tension is particularly significant in the area of digital technology: policy-makers often struggle to fully understand the issues, and problem definition is subject to a high degree of political contestation enabled by digital technology itself. The talk discussed these questions and illustrated them through three specific analyses: (1) the emergence of content moderation as a political issue, (2) the effects of decentralised social media on user sharing behavior, and (3) the role of media coverage for platform policy change.

There are no simple solutions to complex societal challenges. Whether it is climate change or dealing with the devastating impacts of the coronavirus pandemic, the questions these problems pose to humanity have no single correct answer. Addressing them requires the collaboration of governments, charities, businesses and individuals. However, at a time when society seems to be dominated by dogma and discord, building consensus on what action to take may seem like the biggest hurdle to overcome. We cannot overlook the role of the Internet in fomenting divisions. Fake news and social media bubbles filter our reality and have the power to entrench us on one side of the argument, preventing us from understanding the views of others. However, research on Collective Intelligence also notes that technology can be a powerful tool to help us find common ground, even in cases where it seems we could not be further apart.

For more than a decade, I have argued for a new kind of collective intelligence, elsewhere referred to as Contested Collective Intelligence (CCI), which is mediated by new technologies for dialogue and argumentation and specifically aims to help people make sense of and co-create innovative solutions to complicated challenges. The CCI tools that we have developed harness the power of technology to enable people around the world to construct shared understandings, even when, at first glance, they disagree. Combining advanced computational methods, such as natural language processing and machine learning, with intuitive multimedia interactions, these user-friendly tools harness and structure online conversations to identify stated and unstated points of agreement within a discussion group and in this way help the group better understand and address complex problems.

In this talk, I provided a brief overview of Contested Collective Intelligence research and presented two of our CCI tools: bcause.app and democraticreflection.org.

strategicDecidim (decidim.org) is a digital platform for participatory democracy, built entirely and collaboratively as free open source software. More specifically, Decidim is a web environment that allows anyone to create and configure a technopolitical network. The platform can be deployed by any organization (local/regional/national governments, universities, nonprofits) to host large-scale citizen participatory processes for strategic planning, participatory budgeting, public consultations and collaborative policy-making. The project was launched in 2017 in Barcelona and, 5 years later, there are hundreds of active instances around the world. In this talk, key lessons from research and practice with Decidim were shared, including the impact of its deliberative platform design and the technopolitical principles that guide the participatory development of the project.

This talk was organized in three parts. The first part revolved around communication studies and the calls for a need to re-define core definitions and operationalizations of communication studies, so that the pivotal role of AI technologies is acknowledged and examined. Towards this end, there is increasing discussion about examining the agentic role of AI, in addition to the traditional anthropocentric conceptualization of the study of communication. In the second part of my talk, I presented some examples from a recently concluded work from the Horizon 2020 project Co-Inform (Co-Creating Misinformation Resilient Societies, proposal 770302). Our focus in this work was on investigating how a co-created browser plugin influenced citizens’ perception of misinformation and their subsequent actions. As part of this talk, I also briefly discussed the media and information literacy implications of such work for learning, education, and the design of such interventions. I concluded the talk with a summary of open areas of inquiry, informed by the Co-Inform work on how to combat misinformation on social media.

Democracy is more than a decision by the majority. It is also the rule of law, encompassing fundamental rights such as freedom of expression, property, access to courts etc. on the individual level and institutions with defined competencies such as parliament, government etc. on the institutional level. Democracy is also a process. The involvement of citizens in the legislative process, the transparency of the debate as well as on financing are essential for the functioning of a modern democracy. Switzerland has opted for quite a radical system allowing popular initiatives to amend more or less any article of the Swiss constitution.

Access to large quantities of data, artificial intelligence or both elements combined may fundamentally influence democracy. They may enhance the consultation process initiated from the authorities as well as bring together like-minded people to draft and launch a popular initiative. These possibilities are still unexplored both by private and state actors.

The successful design of digital democracy systems presents a multidisciplinary research challenge. In this short presentation, I explained what computational social choice is and I argued why tools and techniques from this field are relevant for the design of online decision-making platforms and other digital democracy systems.

Recommender systems and social networks are often faulted to be the cause for creating Echo Chambers – settings where people mostly encounter news that match their preferences or those that are popular among similar users, resulting in their isolation inside familiar but insulated information silos. Echo chambers, in turn, have been attributed to be one cause for the polarization of society, which leads to the increased difficulty to promote tolerance, build consensus, and forge compromises.

To escape these echo chambers, we propose to change the focus of recommender systems from optimizing prediction accuracy only to considering measures for social cohesion. The talk also succinctly presented some results from an empirical study investigating if such a recommender system would actually have the desired results (see [1] for details).

We discussed possible audiences for publications inspired by the seminar. These include policymakers, funding agencies, research evaluators, academics, citizens (in general) and activists.⁶⁶6Representatives of organisations active in promoting higher citizen involvement in democratic decision-making had been invited to the seminar but were unfortunately unable to join. We recommended that addressing policymakers and academics would have the higher priority.

The goals we discussed were aimed at the stages of the democratic process for which (computational) support could be provided. At a high level, the goals we identified were:

1. 1.

   Citizens

   • $\mathrm{■}$

     Inform, as a basis for forming an opinion

   • $\mathrm{■}$

     Ensure that citizens (and other actors) have appropriate and verifiable information for their decision making

   • $\mathrm{■}$

     Deliberate, as a basis for forming an opinion

   • $\mathrm{■}$

     Ensure that citizens have sufficient opportunities to expose themselves to different opinions and deliberations (incentives for) engagement (with both politicians and administrations/civil servants)

2. 2.

   Procedural/Institutional

   • $\mathrm{■}$

     Improve the democratic processes through appropriate technology

   • $\mathrm{■}$

     Protect individuals from repercussions due to their political activity

In addition to developing the goals of technological support for democratic processes, we identified the common goals with a complementary initiative: the “Vienna Manifesto on Digital Humanism”.⁷⁷7Vienna Manifesto on Digital Humanism https://dighum.ec.tuwien.ac.at/dighum-manifesto/

With floundering trust in politics and declining electoral participation rates in many of the world’s democracies, the call for efforts to re-engage citizens grows ever louder. Over the past two decades, this has resulted in a large number of different participatory formats, including participatory budgeting, citizen juries, and minipublics. In nearly all cases, the number of participants in those formats has been limited. Digitalization offers new opportunities to create large-scale deliberations and co-creation projects for citizens. Scale is important because public acceptance of new participatory forms is enhanced when large segments of the public are represented. There have already been excellent experiments with large-scale technnology-aided deliberation (see the keynote by Pablo Aragón).

Just because one builds platforms, however, does not mean that all problems of participation are resolved. It is very clear that unequal participation from different groups in society remains an urgent problem. These problems start with recruitment but do not end there. Citizens may not be active participants, they may not be heard because they are less well equipped expressing their views, and they may ultimately drop out, perhaps more frustrated than they started. A major challenge in the digital age is to engage citizens, to retain them, and to ensure that they find an effective voice in deliberative and co-creative platforms. We think here of automatic moderation, which detects when certain individuals have fallen silent or have made inputs that have gone unanswered. We can also think of assistive technologies that help citizens sort and formulate their ideas and arguments in order to enhance their impact.

While citizens are key actors in democratic society, we cannot lose sight of other actors, including political parties, representatives, and bureaucrats. We still find that parties and representatives have a tenuous grip of digital technology. Opportunities are missed and risks under-estimated. Enhancing digital skills within the representative and executive institutions is essential if democracy is to flourish in the digital age. Here, the principles laid out in the goals should be a guiding light.

The group discussed the relevance of narratives about digital technology and democracy, and how they are lined to perceived dangers such as misinformation, echo chambers, and the reproduction of bias through algorithmic decision-making. Importantly, there may be a significant mismatch between the discourse around those dangers (e.g. in the media) and their actual importance. An example is “fake news,” which were widely discussed as a main driver of political events such as Brexit and the election of Donald Trump. Verification of those claims by independent researchers was initially very difficult due to problems of access to social media data, which platforms (in particular Facebook) obstructed with arguments linked to privacy and data protection. When reliable research findings became available, they tended to disconfirm many of the ideas around fake news. For example, several independent studies, using different data and methods, concluded that the consumption and sharing of fake news is – on average – very low. Instead, they are concentrated among a specific subgroup of people who tend to be older and very conservative. While these findings do not imply that fake news are not a problem, they point to different policy responses than the initial narrative did.

Against this background, the group discussed specific issues driving the power of narratives about digital technology and democracy and inhibiting research that could challenge them, namely, the ground truth problem. The ground truth problem means that agreeing on categories (e.g., what is “fake news”?) is inherently contested, which makes it very difficult to monitor them. Moreover, there is often a denominator problem, that is, the lack of a benchmark to assess the prevalence of behaviors. By contrast, the media have no trouble finding examples: there is a lot of everything on the internet. Furthermore, both ground truth and denominators are moving targets, because of potentially rapid change in underlying problems (e.g., COVID and the Ukraine war) as well as the relevance and user base of different platforms (e.g., the rise of TikTok).

The group discussion concluded with a brainstorming session regarding possible solutions as broader recommendations.

A common way to break down the policy life cycle results in the following five stages:

1. 1.

   Agenda setting

2. 2.

   Policy formulation

3. 3.

   Decision making

4. 4.

   Policy implementation

5. 5.

   Policy evaluation

Usually, this process is a loop, as after policy evaluation the policy may be terminated and a new agenda might be set or discussion about the existing policy may continue. The reader may note that the process does not need to be strictly sequential and stages may be skipped or repeated.

The working group discussed whether and how the digitalization of democracy means changing these existing stages by rethinking and restructuring process steps, or whether the overall process is stable and the speed and the agility of the process change. We addressed this question by discussing the past, the present and the desired future execution of policy life cycle stages as follows:

Considering stage 1, digitalization has profoundly changed the means for agenda setting. In the past, media gatekeepers have hugely influenced agenda setting and were not easily swayed by activists. Nowadays, traditional media still play a major role, but social media enlarge the set of voices that are heard, which may help minorities, but which increasingly often leads to information overload of the public. As a consequence, few topics are high on the agenda and many lower ranked topics tend to be forgotten. Specific tools like Decidim may help to spread the valuation of importance more evenly among a broader range of topics and, hence, their regular use should help in the future. A core objective for the future would be to “make uncertainty productive”. Current discussions often lead to overly simplistic solutions, where only few aspects are considered because of information overload. If future tools and processes could help to manage a multiplicity of uncertain dimensions, agenda items could be managed in ways that would improve the joint satisfaction of multiple goals.

Considering stage 2, in the past policy formulation could be characterized by tight control by executive and legislative powers who oversaw control of relevant knowledge. Political parties and lobby groups were highly active and only in a minority of cases activism and street protests would contribute. Today, there is a tendency towards increased transparency. Social media-supported activism has led to new kinds of protests that increased the number of actors, while political parties exhibit declining numbers of members. Because of digitalization, knowledge can and is more broadly shared allowing the broad public to know about details of formulated policies.

With regard to the future, we argue in favor of increased transparency about lobbyism and specifically transparency of the policy formulation process. Because of increased complexity of regulations, it is not only important what is written, but who wrote it. Overseeing implications of policies seems only possible by stronger collaborations between politicians and professionals in order to achieve evidence-based policy formulation. Many questions remain open about such co-design of policies: who has access? Who has the right to be heard by their representative? How is consultation fed back to the public commenters?

Considering stage 3, decision-making may change more slowly than other aspects of the policy life cycle. Thereby, the success of novel means of e-voting or liquid democracy does not only depend on the process, but on the assumptions underlying this process. When voting is strictly confidential (as it is in Germany), there does not seem to exist a technology that makes e-voting as sound proof as traditional voting with paper ballots. This is different, if votes can be made public, because then it becomes easy to monitor the soundness of an electronic voting process.

Considering stage 4, in the past policy implementation focused on guiding people by giving or taking money and imposing court orders. Today, we observe a digital turn in connection with a behavioral turn trying to nudge people using behavioral politics and behavioral economics or appealing to citizens as an alternative or complementary means to imposing laws.

For the future, we would wish for governmental processes to be streamlined by digitalization for efficiency and effectiveness as well as an administration that seamlessly serves the citizens rather than imposing heavy administrative burdens on citizens and organizations. A key ingredient to such support may be the real-time observation of individuals and organizations, which, however, bears the dangers of ubiquitous surveillance and manipulation, as they have become obvious in experiments with social credit systems. It remains an open question how to take advantage of benefits while avoiding these heavy drawbacks – and whether this is possible at all.

Little progress has been seen regarding stage 5, policy evaluation. In the future, we might perhaps see more comprehensive consultation and feedback by the public. It would be desirable to formulate key performance indicators that can be reported automatically already during policy formulation. Also, open data might help to judge appropriateness of policies.

Finally, these structures need to be understood and elaborated in terms of overarching or meta perspectives of these process stages. As illustrated by the overarching objective of (non-)anonymity of voting, these meta perspectives may deeply affect the working of the digitalization-enhanced policy life cycle.

Within the technology group, we first examined what technology can do for democracy. We tried to ask ourselves which democratic challenges we think are particularly well suited to be addressed by a technology-mediated approach, and we asked what are the main challenges and opportunities we see emerging in the near future in the field of technology research for democracy. These challenges and opportunities are summarised below. We propose them as key points of focus for the fields of computer, social and political sciences interested in the study, design and application of technology for democracy.